In recent years IP SANs have become increasingly more pervasive in enterprise and SMB environments.  IP SANs enable the centralisation of storage and enable administrators to provide and manage scalable, fault-tolerant and redundant storage from a single access point. Several open source projects enable any startup to build their own IP SAN such as Openfiler, Open-E etc. Large enterprise level IP SAN providers include EMC and NetAPP. This  technology almost exclusively is based on the iSCSI protocol.  Basically iSCSI enables the transfer of SCSI commands over IP thus enabling a computer to interact with storage over a LAN. Because SCSI commands interface at the block level with a disk, they offer significantly more performance that file-based shared storage protocols like SMB and NFS. Unlike file-based shared storage protocols however, iSCSI “targets” (shared iscsi volumes) can’t be shared amongst multiple computers. They are however ideal for VM disks and dedicated disks for thin clients. At my workplace we use iSCSI extensively.  Several enterprise applications rely on iSCSI. For instance, developing a web application application with active/passive failover (although can be implement without iSCSI) is more efficient with iSCSI. Moreover, enterprise hardware such as Tape Libraries now rely on iSCSI to enable access to these hardware over a LAN.

I recently evaluated various architectures for deploying a failover clustered web application. Open source projects like DRBD  enable the deployment of such a cluster however they solve the problem in an inefficient manner. DRBD essentially requires the use of two disks on both servers that will constitute the  failover cluster. Anything written to the disk is replicated to the other nodes, and if the active server fails, another server can take over and resume serving the application. Because of the replication DRBD causes quite a bit of network traffic. Windows Server 2008 R2 solves the problem in a more efficient manner. Windows failover clustering  relies on an iSCSI shared storage. All application data is written to the shared storage, and if the server fails, the iSCSI storage is simply mounted on the failover node and the application resumes hosting. This is more efficient because the data is written only once, rather than being replicated.

An iSCSI initiator is a software that enables a computer to “mount” a remote iSCSI share (called a target). Microsoft has shipped an iSCSI initiator with Windows, Linux has open-iscsi, if Apple really wants a chunk of enterprise computing and be relevant to cloud computing enabled environments it needs to ship iSCSI initiators with Mac OSX. The only option available for users like me are commercial initiators like gloablSAN ($89).

Last Week’s Economist issue contained a series of articles on how technological innovation has shifted towards consumer technology. The argument generally was, that in the past innovation at the consumer space always lagged technological innovation at the government and military spaces. But that might be changing and might have already done so. Amazon offered cloud computing to customers before companies started targeting enterprises and governments. Similarly, the latest smartphones lead in terms of technology compared to smartphones prevalent at  large enterprises and governments (Blackberry anyone?). Today I came across this article at Fast Company that lays out the “War” leading technology companies will be engaging in to make significant inroads into the consumer space. Conspicuously absent from this list are companies that focus on large enterprises like IBM, HP, Dell and even traditionally consumer oriented companies like Dell.

I’m personally excited by the focus on the consumer, as by putting more powerful technology at the hands of the consumer countries may become more productive.

I finally got my hands on an iPhone 4S. I upgraded my iPhone 4 contract at the O2 Store in Cribbs Causeway here in Bristol on Saturday. Fortunately, I didn’t have to queue very long to get one!

Before buying it, I headed off to the Apple Store to check Siri out. I was interested in checking it out before I bought the phone, because I wanted to evaluate if it can pick up my accent. I’m not a native english speaker, therefore, several voice recognition engines in the past have failed for me. I was surprised that Siri worked a 100% for me.

I’ve been experimenting with Siri since Saturday, and I have to say that mostly it has impressed me. Although, here in the UK, the capabilities of Siri are slightly more limited than in the US. It doesn’t do anything related to Maps, business and address searches. Also, I’m surprised that it doesn’t understand the word “tweet”, “iMessage” and so on (Its surprising since both technologies are part of the iOS 5, so at least it should understand that). Nevertheless since its “beta” I’ll ignore these shortcomings for now.

The beauty of Siri for me, is the capability of translating human speech into API calls. (e.g. “Do I need an umbrella” translates to a query for checking the weather, remind me to fix this code when I get to work translates into a geotagged reminder etc..). I’ve been thinking about the implications of this and the they are profound indeed. I’m sure, Apple, is considering of opening up Siri to third party application developers. I look forward to saying to Siri “calculate me a route between Paddington to Knightbridge via the underground”, after which it should open a Tube Underground app and have to route plotted for me. Similarly, if I ask when is the next bus at a particular station, it should give me a schedule of the upcoming buses at that particular bus station. All this kind of integration does require help from third parties however, I can’t envision how such integration would be achieved, as the potential for conflicts is significant.

The situation of where Siri is going in future will certainly be more clear by the time the next iPhone is announced. One thing, I don’t get however, is why Siri is not available on other iOS devices. One reason touted is that it requires the extra processing power available on the iPhone 4S. However, I’m not exactly sure thats true, since I’ve seen that its response time is directly proportional to my internet connectivity. If my phone is in airplane mode for example I get this:

Apparently Siri sends off data to some server for remote processing before it formulates its response. Therefore, the need for extra local processing power is mitigated. So, why is there no Siri for other iOS devices?

Learning about Probability and Statistics in my opinion is as important, if not more, to learn than any programming paradigm out there. Programmers and computer scientists in general love to delve deep into software and computer systems. This often involves the analysis of lots of data, in order to make sense of the data, descriptive statistics are vital.  Moreover, several fields of Computer Science now including Artificial Intelligence and Machine Learning feature methods that are based primarily on statistical approaches.

So whats a good way learn about this domain? Today I found a sufficiently comprehensive and accessible online resource.

Since the last few months someone or some people have trying to create a Facebook account using my gmail address. I have received countless “Welcome to Facebook” mails. Fortunately, Facebook does provide a mechanism form where you can block the creation of accounts using a specific emails address. The problem is… it doesn’t work.

It always begins with an email like this:

Along with this email, Facebook sends an email verification email, which specifies  link for reporting unauthorised account signups with your email address.

I report it and I get this..

However, the email has clearly not been disabled as I keep getting account signups!

Recently there is has been lots of innovation in App design for tablets. Apps like Qwiki, Flipboard, Zite, Pulse, Reedr have really figured out how to give a good user experience on a tablet. However, sadly, most twitter clients have not. Even iphone twitter clients like TweetBot are better than most iPad twitter clients as they are more optimised for the display size. The fundamental problem seems to be that twitter client developers think that iPad is just a scaled up iPhone.

With the help of this thinking various successful iPhone twitter client developers have released new iPad clients which are nothing more than scaled up versions of their iPhone application, with some added iPad elements. For instance, the following screenshots are from Twitterific and Echofon on the iPad.

T

As you may observe, over 90% of the screen it taken up by a single UITableView. IMO the UITableView is an efficient way to present of a list of items (or a stream of tweets) in a small display. It makes no sense however, to just adopt the same thinking on a larger screen. A better use of the iPad screen is made in TweetDeck for instance. It makes more rich use of the screen, as the following screenshot depicts. However, unfortunately, the TweetDeck app has some performance issues.

Recently a new bread of applications such as Flipboard and Zite have emerged. These applications enable you to consume information in a magazine like fashion. I believe they too are an efficient way to consume your twitter feed and lists however they are not design to be full fledged clients. I believe there is a gap in the iPad App market for twitter clients that make rich use of the screen yet do so in a computationally efficient manner.

If you are aware of any good twitter clients please post to the comments…

I’ve been using a lot of VMs recently. By default the number of loopback devices set in Ubuntu is just 8. So if you want to mount 10 VM disks, your out of luck.

An easy way I’v found to increase loopback devices is:

edit /etc/modprobe.d/options

and enter the line (replace n, for the number of loopback devices you want. Max. number is 255).

options loop max_loop=n

Last week my Time Machine Backup Hard disk suddenly died! Everytime I plugged the hard disk, my mac did not recognise the file system (it was a HFS+ partition) and asked my to format it. I did format it, only to realise that I had lost some precious data.

How do you recover data from a formatted hard disk? Enter Data Rescue from PROSoft Engineering. I used Data Rescue II (only to realise afterwards that a new version was available). It took more than 2 days to scan the hard disk (1 TB hard disk, sector by sector analysis) and afterwards it recreated the files it found.

Recreation of the files, was also a lengthy process, and took nearly 10 hrs. After it recovers the files it presents a list of the kinds of files you might be interested in restoring. I selected the files I was interested in, research papers and my iPhoto Collection.

Recovery of the selected files (around 55.6GB) took around 3 hrs.

All in all I’m really grateful for such fantastic software. For all the windows users out there…. My NTFS hard disk has failed as well (bad start to the year! ). Currently I’m using ParetoLogic’s Data Recovery Pro. The data recovery process is currently on going (since 5 days).